In B. Sapp, A. Toshev, and B. Taskar. Cascaded models for articulated pose estimation. In ECCV, 2010; F. Fleuret and D. Geman. Coarse-to-ne face detection. Int. J. Comput. Vision, 41:85 107, 2001; J. B. D. Munoz and M. Hebert. Stacked hierarchical labeling. In ECCV, 2010.; and D. Weiss and B. Taskar. Structured prediction cascades. In Proc. AISTATS, 2010, the authors improve performance of certain computer vision tasks (viz. pose estimation, face recognition, etc.) using coarser information. In P. Slay. Coarse-to-ne natural language processing. Ph.D Thesis UC Berkeley, 2009., coarser information is used to improve the performance of neuro-linguistic programming (NLP) models. The existing work, however, mainly considers or uses the coarser information in classification and the structured prediction setting. No work is known that uses the coarser information for the regression setting.